1. 1 Ronald C. Kessler, Ph.D. Department of Health Care Policy Harvard Medical School March 6, 2008 Comorbidity of Anxiety Disorders in the National Comorbidity Survey Adolescent Supplement (NCS-A)

2. 2 NCS-A study design Nationally representative sample of n = 10,200 adolescents Sampled from a nationally representative sample of 230 schools All respondents were ages 13 – 17, English- speaking, and students Face-to-face interviews with respondents Self-administered questionnaires with parents

3. 3 The NCS-A instruments Adolescent version of WHO CIDI – 3.0 with adolescents Informant version of CIDI with parents School survey focused on mental health resources with school Principals and mental health coordinators

4. 4 DSM-IV anxiety disorders assessed Panic disorder with or without agoraphobia Agoraphobia without a history of panic disorder Specific phobia Social phobia Generalized anxiety disorder Post-traumatic stress disorder Obsessive-compulsive disorder Separation anxiety disorder

5. 5 Other DSM-IV disorders assessed Mood disorders (MDD, dysthymic disorder, BPD) Externalizing disorders (ADHD, ODD, CD, IED, eating disorders) Substance disorders (alcohol and drug abuse- dependence)

6. 6 Concordance of DSM-IV diagnoses based on the CIDI and the K-SADS Panic disorder with or without agoraphobia74.499.20.87 Agoraphobia without a history of panic disorder81.998.70.90 Specific phobia96.992.10.94 Social phobia65.695.80.81 Generalized anxiety disorder60.199.30.80 Post-traumatic stress disorder59.998.00.79 Sens Spec AUC

7. 7 DSM-IV/CIDI lifetime prevalence estimates of anxiety disorders Panic disorder with or without agoraphobia 2.3(0.2) Agoraphobia without a history of panic disorder 2.4(0.2) Specific phobia19.3(0.8) Social phobia 9.1(0.4) Generalized anxiety disorder 2.2(0.4) Post-traumatic stress disorder 5.0(0.3) Separation anxiety disorder 7.6(0.3) % (se)

8. 8 Age of onset distributions for anxiety disorders

9. 9 Age of onset distributions for mood disorders

10. 10 Age of onset distributions for externalizing disorders

11. 11 Age of onset distributions for substance disorders

12. 12 Age of onset distributions for each class of disorders

13. 13 Rotated (promax) factor loadings (standardized regression coefficients) of lifetime DSM-IV/CIDI diagnoses at the level of the person-year

14. 14 Rotated (promax) factor loadings (standardized regression coefficients) of lifetime DSM-IV/CIDI diagnoses at the level of the person-year

15. 15 Discrete-time survival analysis of associations between temporally primary disorders and the subsequent onset of secondary disorders Person-year data array

16. 16 Discrete-time survival analysis of associations between temporally primary disorders and the subsequent onset of secondary disorders Person-year data array Temporally primary disorders are treated as time- varying covariates

17. 17 Discrete-time survival analysis of associations between temporally primary disorders and the subsequent onset of secondary disorders Person-year data array Temporally primary disorders are treated as time- varying covariates A series of models was estimated to evaluate the effects of primary disorders on onset of secondary disorders

18. 18 Survival analysis summary of results Everything predicts everything in bivariate models

19. 19 Survival analysis summary of results Everything predicts everything in bivariate models Many sign flips in additive multivariate models

20. 20 Survival analysis summary of results Everything predicts everything in bivariate models Many sign flips in additive multivariate models Marginal effects stabilize in a simple global interactions model

21. 21 Survival analysis summary of results Everything predicts everything in bivariate models Many sign flips in additive multivariate models Marginal effects stabilize in a simple global interactions model Global interactions are significant and consistently sub-additive

22. 22 Survival analysis summary of results Everything predicts everything in bivariate models Many sign flips in additive multivariate models Marginal effects stabilize in a simple global interactions model Global interactions are significant and consistently sub-additive More complex interaction models find no evidence of domain-specific effects

23. 23 Survival analysis summary of results Everything predicts everything in bivariate models Many sign flips in additive multivariate models Marginal effects stabilize in a simple global interactions model Global interactions are significant and consistently sub-additive More complex interaction models find no evidence of domain-specific effects Marginal effects show some evidence of domain specificity, but domain-specific effects do not account for all significant associations

24. 24 Survival analysis summary of results regarding marginal effects There are 240 logically possible time-lagged associations among 16 disorders

25. 25 Survival analysis summary of results regarding marginal effects There are 240 logically possible time-lagged associations among 16 disorders But some of these do not exist by definition (e.g., BPD predicting MDD)

26. 26 Survival analysis summary of results regarding marginal effects There are 240 logically possible time-lagged associations among 16 disorders But some of these do not exist by definition (e.g., BPD predicting MDD) Others had too few cases for analysis (e.g., drug disorders predicting ADHD, as onset of ADHD occurs so much earlier than onset of drug abuse)

27. 27 Survival analysis summary of results regarding marginal effects There are 240 logically possible time-lagged associations among 16 disorders But some of these do not exist by definition (e.g., BPD predicting MDD) Others had too few cases for analysis (e.g., drug disorders predicting ADHD, as onset of ADHD occurs so much earlier than onset of drug abuse) As a result, we had a total of 236 time-lagged associations to study

28. 28 Survival analysis summary of results regarding marginal effects 91.5% of the 236 survival coefficients were positive

29. 29 Survival analysis summary of results regarding marginal effects 91.5% of the 236 survival coefficients were positive 58.8% of the positive coefficients were statistically significant at the.05 level (two-sided tests)

30. 30 Survival analysis summary of results regarding marginal effects 91.5% of the 236 survival coefficients were positive 58.8% of the positive coefficients were statistically significant at the.05 level (two-sided tests) None of the negative coefficients was statistically significant

31. 31 Survival analysis summary of results regarding marginal effects 91.5% of the 236 survival coefficients were positive 58.8% of the positive coefficients were statistically significant at the.05 level (two-sided tests) None of the negative coefficients was statistically significant The within-domain coefficients were generally larger than the between-domain coefficients

32. 32 Mean values (and percent statistically significant at the.05 level, two-sided test) of marginal effect survival coefficients within and between domains Fear Distress Externalizing Substance Other Fear1.5 (83)0.8 (44)1.0 (75)0.4 (0)0.8 (75) Distress1.5 (75)1.5 (84)1.3 (67)0.5 (17)0.7 (63) Externalizing0.9 (33)0.6 (33)2.0 (100)0.7 (67)0.6 (50) Substance0.7 (29)0.6 (17)1.1 (40)3.0 (100)0.3 (25) Other1.3 (82)0.4 (45)1.0 (75)0.4 (38)0.5 (25)

33. 33 Mean values (and percent statistically significant at the.05 level, two-sided test) of marginal effect survival coefficients within and between domains Fear Distress Externalizing Substance Other Fear 1.5 0.8 1.0 0.4 0.8 Distress 1.5 1.3 0.5 0.7 Externalizing 0.9 0.6 2.0 0.7 0.6 Substance 0.7 0.6 1.1 3.0 0.3 Other 1.3 0.4 1.0 0.4 0.5

34. 34 Mean values of substantively significant odds-ratios within and between domains Fear Distress Externalizing Substance Other Fear 4.52.22.7-2.2 Distress 4.5 3.7-2.0 Externalizing 2.5-7.42.0- Substance 2.0-3.020.0- Other 3.7-2.7--

35. 35 Conclusions It is not entirely clear that it makes sense to speak of a single “domain” of disorders known as “anxiety disorders.” Distinct fear and distress domains clearly exist. The situation with OCD might be part of yet a third domain.

36. 36 Conclusions The Fear and Distress Disorders Association of America

37. 37 Conclusions FD 2 A 2

38. 38 Conclusions It is not entirely clear that it makes sense to speak of a single “domain” of disorders known as “anxiety disorders.” Distinct fear and distress domains clearly exist. The situation with OCD might be part of yet a third domain. Fear disorders are the most commonly occurring early-onset mental disorders.

39. 39 Conclusions It is not entirely clear that it makes sense to speak of a single “domain” of disorders known as “anxiety disorders.” Distinct fear and distress domains clearly exist. The situation with OCD might be part of yet a third domain. Fear disorders are the most commonly occurring early-onset mental disorders. Fear disorders have a very strong pattern of cumulation over time, with the onset of the first strongly predicting the subsequent onset of the second, the second predicting the third, and so forth.

40. 40 Conclusions (cont.) Fear disorders also strongly predict the subsequent onset of a wide range of other disorders, the main exception being substance disorders.

41. 41 Conclusions (cont.) Fear disorders also strongly predict the subsequent onset of a wide range of other disorders, the main exception being substance disorders. Social and specific phobias are as important here as panic disorder. This is striking in light of the general perception that child-adolescent phobias are not very “important” in comparison to other commonly occurring early-onset disorders.

42. 42 Conclusions (cont.) Fear disorders also strongly predict the subsequent onset of a wide range of other disorders, the main exception being substance disorders. Social and specific phobias are as important here as panic disorder. This is striking in light of the general perception that child-adolescent phobias are not very “important” in comparison to other commonly occurring early-onset disorders. It’s not clear that these associations are causal.

43. 43 Conclusions (cont.) An impediment to addressing the causality issue is that early-onset fear disorders are under-treated.

44. 44 Conclusions (cont.) An impediment to addressing the causality issue is that early-onset fear disorders are under-treated. We need effectiveness trials that evaluate the effects of timely detection and treatment of early-onset fear disorders on the subsequent onset of other mental disorders.

45. 45 Conclusions (cont.) Distress disorders, which include not only depression but also GAD and PTSD, typically have later ages of onset.

46. 46 Conclusions (cont.) Distress disorders, which include not only depression but also GAD and PTSD, typically have later ages of onset. We know that distress disorders are seriously impairing in their own right and greatly increase the severity of comorbid disorders.

47. 47 Conclusions (cont.) Distress disorders, which include not only depression but also GAD and PTSD, typically have later ages of onset. We know that distress disorders are seriously impairing in their own right and greatly increase the severity of comorbid disorders. It is especially important in light of these other results that distress disorders are quite important in predicting the subsequent onset of diverse secondary disorders.

48. 48 Conclusions (cont.) As a result, distress disorders are usually highly comorbid. The vast majority of people with all major distress disorders have a history of other mental disorders.

49. 49 Conclusions (cont.) As a result, distress disorders are usually highly comorbid. The vast majority of people with all major distress disorders have a history of other mental disorders. As with fear disorders, it’s difficult to know what causes what in comorbidities involving distress disorders.

50. 50 Conclusions (cont.) An added complication in sorting out causal priorities for distress disorders is that distress disorders have a very protracted risk window compared to fear disorders, as indicated by the wider IQR of the AOO distribution (close to 30 years for distress vs. 7-10 years for fear disorders).

51. 51 Conclusions (cont.) An added complication in sorting out causal priorities for distress disorders is that distress disorders have a very protracted risk window compared to fear disorders, as indicated by the wider IQR of the AOO distribution (close to 30 years for distress vs. 7-10 years for fear disorders). Because of this, prospects for getting insights into complex causal connections based on targeted intervention and follow-up are much greater for fear than distress disorders, making the intervention experiments noted above an especially important research agenda for the future.

52. 52 www.hcp.med.harvard.edu/ncs